topic Re: default route problem in Operating System - HP-UX

default route problem

Josef Jirkovsky — Mon, 03 Mar 2003 13:51:05 GMT

Hi everybody,
I am installing an L1000 (9000/800) with HPUX
11.00 and i am faced with a strange (for me)
routing problem.

My network topology is the following:
the L1000 is part of a 192.160.1.0 network with address 11. The default gateway is 192.168.1.100
with Internet access.

At first (after reboot), all works perfectly = i can ping
an Internet address without problem. After about
140 pings the default route doesn't work any more.
I can only ping systems on the 192.168.1.0 network.

If i delete the default route (route delete default 192.168.1.100) and then add it again (route add net default 192.168.1.100 1), all is ok for an other 50 pings (the number of successful pings varies).

Did i miss some configuration issue (though it's not my first L1000 i'm installing)?
For now, i don't think it's the network card (i can ping the other systems on the local network).

Any help is much appreciated

josef jirkovsky
TACTIC SA
Paris, France

Re: default route problem

T G Manikandan — Mon, 03 Mar 2003 14:00:44 GMT

To check whether it is a dead gateway just do

ndd -get /dev/ip ip_ire_status

if the output is IRE_GATEWAY DEAD then you can disable the dead gateway detection by modifying

ndd -get /dev/ip ip_ire_gw_probe

If the output is 1 then it means that the dead gateway detection is turned on.

you can turn it off using

ndd -set /dev/ip ip_ire_gw_probe 0

To make it permanent make entries in the /etc/rc.config.d/nddconf file

Thanks

Re: default route problem

Ross Zubritski — Mon, 03 Mar 2003 14:00:57 GMT

Hi Jolsef,

Is the defualt route in /etc/rc.config.d/netconf?

Thanks

RZ

Re: default route problem

Michael Steele_2 — Mon, 03 Mar 2003 14:02:16 GMT

I've seen this happen with a runaway inetd process, use top for this.

You may also have a denial of service or faulty network application reporting via icmp packets. Use:

netstat -p icmp

-and-

netstat -f inet | grep -i -e drop -e collision -e router -e unreachable

Get a copy of tcpdump to see where the source of the icmp's are, if this is the case.

Re: default route problem

Michael Steele_2 — Mon, 03 Mar 2003 14:05:32 GMT

If dead gateway, which will occur in a DMZ, then no pings will come. Above states: "...After about
140 pings the default route doesn't work any more..."

Re: default route problem

Michael Steele_2 — Mon, 03 Mar 2003 14:20:09 GMT

Check /etc/rc.log for errors. In particular, something like NFS tries but fails to start up.

Re: default route problem

Ron Kinner — Mon, 03 Mar 2003 21:01:31 GMT

T.G. is correct.

It's a dead gateway detection problem. This is a default setting for 11.0. You get about 3 minutes after the initial addition of a gateway before the lack of a ping response from the gateway gets it declared dead.

I bet you cannot ping 192.168.1.100 which is what is causing your problem. A lot of paranoid security people are stopping all replies to pings.

Just do as he says:

ndd -set /dev/ip ip_ire_gw_probe 0

That will fix the problem. Then edit /etc/rc.config.d/nddconf

TRANSPORT_NAME[0]=ip
NDD_NAME[0]=ip_ire_gw_probe
NDD_VALUE[0]=0

Change the [0] to the next higher integer if you already have entries there. That way when you reboot it will stay.

I have seen reports where the ndd change did not work. I suspect there is an ndd patch missing so if it does not work for you then look for the latest ndd patch. I know our 11.0 system (which does not have the ndd patch) does not even show the existance of the option when you do ndd -h.

Ron

Re: default route problem

Josef Jirkovsky — Tue, 04 Mar 2003 08:48:24 GMT

Thank you very much!
You are right, i had a "dead gateway" problem
and setting the detection to 0 was the successful
action to take.

I have to keep up to date and not to stop learning.

have a fine day!
josef